The Vector Grounding Problem

Dimitri Coelho Mollo, Raphaël Millière

Large language models (LLMs) produce seemingly meaningful outputs, yet they are trained on text alone without direct interaction with the world. This leads to a modern variant of t…